Message boards : Theory Application : Odd CPU time for a single CPU core task
Message board moderation

To post messages, you must log in.

AuthorMessage
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 423
Credit: 22,582,008
RAC: 5,678
Message 40281 - Posted: 25 Oct 2019, 12:01:40 UTC
Last modified: 25 Oct 2019, 12:04:49 UTC

https://lhcathome.cern.ch/lhcathome/result.php?resultid=249612757
The above linked task reported a huge CPU time compared to run time. While it was running the discrepancy was even higher than what was recorded at the end. So the task was run using one CPU core (via app_config.xml) and web preferences are set for 4 cores (that is used for Atlas tasks). I noticed the peculiar behavior when the task had been running about 6-7 hours. The CPU time at that point was over 100 hours. This was showing even on the vbox GUI Top (Alt-F3) display the pythia.py was showing over 88 hours of CPU time for 1 CPU job. But it was actually using only 1 CPU, only CPU time measurement had some kind of multiplier on it. Later on it started to behave so that CPU time stopped increasing altogether. You can see this behavior on the stderr.

The computer (win 10) has 16 threads and Boinc is allowed to use 15 of those. There were 15 CPUs in use at the time (2 x 4 CPU Atlas tasks, 2 x Theory, 1 x CPDN and 2 x GPU Seti + 2 x GPU Einstein (each GPU task reserves 1 CPU)). So what could be the reason for this weird CPU time?

[edit]This is a one off task, the rest of Theory tasks that machine has finished look normal.
ID: 40281 · Report as offensive     Reply Quote

Message boards : Theory Application : Odd CPU time for a single CPU core task


©2020 CERN